The Political Economy of Automation: Occupational Automatability and Preferences for Redistribution

Munich Personal RePEc Archive

The Political Economy of Automation:

Occupational Automatability and Preferences for Redistribution

van Hoorn, Andre

Radboud University, Institute for Management Research

2 May 2018

Online at https://mpra.ub.uni-muenchen.de/86460/

MPRA Paper No. 86460, posted 05 May 2018 10:04 UTC

The Political Economy of Automation: Occupational Automatability and Preferences for Redistribution

André van Hoorn

Radboud University, Institute for Management Research, PO Box 9108, 6500 HK, Nijmegen, the Netherlands. Tel: +31 243 611 663, Fax: +31 243 612 379, E-mail: A.van.Hoorn@fm.ru.nl.

The Political Economy of Automation: Occupational Automatability and Preferences for Redistribution

Abstract

Although the importance of technological change for increasing prosperity is undisputed and economists typically deem it unlikely that labor-saving technology causes long-term employment losses, people’s anxiety about automation and its distributive consequences can be an important shaper of economic and social policies. This paper considers the political economy of automation, proposing that individuals in occupations that are more at risk of losing their job to automation have stronger preferences for government redistribution. Analysis of cross-national individual-level survey data from three different sources confirms the effect of occupational automation risk on redistribution preferences. The same effect is found when considering indirect exposure to automation risk through the occupation of one’s spouse or partner and using the automatability of individuals’ own occupation as a generic control variable. In addition, the effect is not limited to the preference for redistribution in general but extends to a preference for a specific policy with redistributive consequences, namely the preference for government support of declining industries.

JEL codes: J23, J24, J31, O14

Keywords: Preferences for redistribution; automation; social insurance; robotization; task content; welfare state; routineness

I. INTRODUCTION

Whereas few people deny the power of technological change to radically transform economies, economists typically emphasize the long-run benefits of creative destruction and technological change for increasing prosperity (Autor 2015; Mokyr et al. 2015). Advances in robotics, task automation, machine learning and artificial intelligence algorithms, however, have been fueling widespread concerns about the possibility of long-lasting technological unemployment and the disruptive effect of labor-saving technology on society (Brynjolfsson and McAfee 2014; Etzioni 2017; Ford 2015; Krugman 2012; Sachs and Kotlikoff 2012; Srnicek and Williams 2015; Summers 2013). The core of such automation anxiety is that technological change has distributional effects between groups in society, meaning that the benefits of technological change are not spread evenly.¹ Recent debate and anxiety thereby highlight a specific automation concern, which is that some workers or occupations end up losing from labor-saving technology not just compared to other groups but also in absolute terms. If large groups in society are indeed in danger of long-run losses from automation this would be a powerful shaper of economic and social policies. Jobs that are more automatable come with higher unemployment risk and economic uncertainty, stoking up demand for government action that mitigates or insures against income losses.

Different estimates of the labor market effects of automation exist. Frey and Osborne (2017), for example, find that approximately 47% of the U.S. labor force is at risk of displacement by machines. A report by McKinsey (2017) similarly finds that roughly six out of ten occupations in the U.S. involves tasks that can be automated for at least 30%. Finally, Acemoglu and Restrepo (2017) consider the effect of robotization on the U.S. labor market, estimating that one additional robot per 1,000 workers reduces the

1 Concern with automation has a long history involving the likes of Keynes (1930) and Marx (1887) before him as well as the famous Luddite movement, which protested automation by destroying weaving equipment in early 19^th century England (Akst 2013; Hobsbawm 1952; Sale 1995). I use the terms automation and labor-saving technology interchangeably. Formally, however, the latter includes both automation and mechanization.

employment ratio by 0.18-0.34 percentage points and wages by 0.25-0.5%.

This paper considers the political economy of automation, linking the automatability of work, specifically individuals’ risk of losing their job to automation, to preferences for redistribution.

Redistributing material wealth has traditionally been one of the most important roles played by governments and individuals’ preferences for redistribution have been widely studied (Ashok et al. 2015;

Corneo and Grüner 2000; Dahlberg et al. 2012; Fong 2001; Kuziemko et al. 2015; Luttmer 2001; Luttmer and Singhal 2011). The workhorse model for understanding differences in preferences for redistribution comes from Meltzer and Richards’ (1981) seminal work, which emphasizes the role of individuals’

income (relative to the median) (see, also, Romer 1975 and Roberts 1977). However, expectations concerning future income or wealth and socioeconomic mobility are also commonly proposed as determinants of redistribution preferences (Bénabou and Ok 2001; Piketty 1995) and confirmed by empirical evidence. Ravallion and Lokshin (2000), for instance, find that negative expectations about future welfare increase demand for redistribution (see, also, for example, Corneo and Grüner 2002). My hypothesis is that greater occupational automation risk leads to a stronger preference for redistribution.

The underlying logic is that individuals are aware of the nature of their job tasks and recognize the implications of their occupational task content for the automatability of their work and hence for their income and wealth prospects and the net pecuniary gains that they can expect from government redistribution.

In order to analyze the effect of exposure to automation risk on preferences for redistribution, this paper follows earlier work on occupational task content (e.g., Acemoglu and Autor 2011) and draws on individual-level survey data to construct an aggregate indicator of the degree of automatability of different occupations. Empirical analysis shows that combining insights on the risk of job loss due to automation from employees actually working in a particular type of job renders a valid measure of occupational automatability or exposure to automation risk. Individual-level cross-national data on self- reported preferences for redistribution and various other factors recognized to affect these preferences come from the International Social Survey Programme modules on Social Inequality (ISSP-SI), four

waves of data in total (ISSP Research Group 2014). The indicator of occupational automatability is matched to these data using codes from the 1988 version of the International Standard Classification of Occupations (ISCO). Seven waves of data from the European Social Survey or ESS (European Social Survey 2016) provide a means for triangulating results and taking into account an alternative set of potential confounders. Similarly, data from the 1996 and 2006 Role of Government module of the ISSP or ISSP-ROG (ISSP Research Group 2008) enables extending results from preferences for redistribution, which are relatively broad and abstract, to preferences concerning a specific policy issue, namely government support of declining industries.

Key challenge for valid identification of the causal effect of occupational automation risk on preferences for distribution is properly controlling for omitted variables. Hence, I consider a large variety of control variables, mostly concerning individual-level differences but also concerning differences between occupations. Rather uniquely, these controls do not only involve standard variables such as education but also, among others, the comprehensive measures of human motivations provided by the basic human values framework, the leading framework of personal values in psychology (e.g., Schwartz 1992; see Figure B.1 in Appendix B). The paper’s main identification strategy, however, is that, instead of considering the effect of individuals’ own automation risk, I consider the effect of individuals’

exposure to automation risk through the occupation of their spouse or partner, while using the automatability of individuals’ own occupation as a generic control variable.

The key findings of this paper are as follows. First, individuals in occupations that are more at risk of job loss due to automation do indeed exhibit stronger preferences for redistribution, indicating an important channel through which automation anxiety can end up affecting societies, independent of automation’s direct labor market effects. Recent technological change seems to increase economic uncertainty for a growing part of the population, prompting citizens to pressure their governments into compensatory social and economic policies. Second, occupational task routineness and complexity do not affect preferences for redistribution when occupational automatability is taken into account. Finally, the effect of occupational automatability on preferences extends to preferences for a concrete economic

policy with redistributive consequences in the form of government support of declining industries.

This paper builds on and contributes to different literatures. The possible economic effects of automation have been widely considered, both theoretically and empirically. However, this work focuses mostly on direct labor market effects and pays little explicit attention to significant indirect effects involving individuals’ attitudes and demand for government policies. Even when the direct labor market effects of labor-saving technology and thus the objective threat of future unemployment and income losses are limited, automation can still have radical effects on society and its policies. The reason is that the subjective perception of automation as an economic threat is already a powerful force influencing people. Automation anxiety affects citizens’ political attitudes, changing the political landscape and pushing governments into action. This paper, then, provides an important complement to studies of the societal implications of key economic trends that focus on direct and objective effects only and pay little attention to indirect but more fundamental changes to society brought on by automation, globalization or other such developments.

Measurement of distinct job characteristics and characterizing the task content of employment is an important topic in different fields, particularly the labor economics literature on job polarization and the new trade literature (Autor and Dorn 2013; Autor et al. 2003; Autor et al. 2015; Blinder and Krueger 2013; Costinot et al. 2011; Goos et al. 2014). The indicator of occupational automatability constructed for the present analysis adds to this literature by providing a valid dimension of occupational task content that complements extant indicators of occupational task routineness and complexity.

This paper resonates most strongly with the literature on preferences for redistribution. The role of net pecuniary gains in shaping individuals’ preferences for redistribution has long been recognized (see Alesina and Giuliano 2011 for a review). Similarly, much attention has been paid to an individual’s income and wealth prospects as determinants of his/her preference for redistribution through their effect on expected costs and benefits of government redistribution. Furthermore, there is a growing body of research that seeks to flesh out the role of income expectations and mobility perceptions in shaping preferences individuals’ expectations about future income or wealth and socioeconomic mobility.

Giuliano and Spilimbergo (2013), for example, find that individuals that have experienced a recession during their formative years end up preferring more government redistribution. Similarly, Alesina et al.

(2018) find that perceptions of intergenerational mobility shape redistribution preferences, even when these perceptions do not accurately match reality. The nature of individuals’ occupation, in contrast, does not yet figure prominently in this literature. The evidence presented in this paper, however, suggests that job characteristics are an important factor shaping individuals’ expectations about future income and socioeconomic mobility.

II. OCCUPATIONAL AUTOMATABILITY II.A. Measuring Occupational Automatability

Measurement of the automatability of occupations in this paper follows the seminal work by Autor et al.

(2003) and other important studies of the task content of employment that have appeared since (e.g., Autor and Dorn 2013; Goos et al. 2014). Two main approaches to assessing the automatability of jobs can be discerned. The first approach revolves around well-established job or task characteristics such as routineness or complexity (Acemoglu and Autor 2011). Following this approach, a job can be seen as comprising a set of tasks, each of which can be more or less susceptible to automation, not least because each task involves more or less routineness and complexity. The automatability of a given job is therefore a function of the automatability of individual job tasks, weighted by the importance of each task as part of the overall content of the job. The second approach is to rely on technology experts judging the overall automatability of different jobs (cf. Frey and Osborne 2017).

The approach in this paper is closer to the latter approach, as it involves aggregated knowledge of employees actually working in a particular type of occupation. The underlying idea is that this approach harnesses the wisdom of crowds and helps overcome biases likely to occur using any of the other two approaches. Calculating occupational automatability on the basis of task content is an intricate process that requires not only that task attributes such as routineness are measured accurately but also that each task is assigned proper weight as part of the overall task content of a given occupation. Expert judgments

face other limitations. A particular concern is the possibility that there are blind spots that are shared by the group of experts asked for their opinion, which is quite likely when the assessment involves a large number of jobs. Two further challenges are lack of cultural diversity among the experts consulted and the use of a system of job classification that does not follow international standards. In these cases, the concern is that the resulting measure of occupational automatability is not universally applicable or valid, likely causing a bias in subsequent empirical analyses.²

To implement my proposed approach, I rely on data collected by the International Social Survey Programme in the 1997 module on Work Orientations or ISSP-WO (ISSP Research Group 1999). This module surveyed respondents from a diverse group of 21 country regions about various aspects of their jobs.³ One specific item asked respondents about the likely effect of automation on employment: “New kinds of technology are being introduced more and more in [country]: computers, robots, and so on. Do you think these new technologies will over the next few years...” The Likert type answer scale provided to respondents comprised five possible answers: “1, Greatly increase the number of jobs?,” “2, Slightly increase the number of jobs?,” “3, Make no difference to the number of jobs?,” “4, Slightly reduce the number of jobs?,” or “5, Greatly reduce the number jobs?” For more than 19,000 respondents, the survey further recorded their occupation using four-digit codes from the 1988 version of the international standard classification of occupations (ISCO). For an additional 5,000 respondents, the 1997 ISSP module

2 As an illustration, in cross-country cross-industry analyses such systematic differences in the accuracy of the industry benchmark considered are known to cause an “amplification bias” (see, for example, Ciccone and Papaioannou 2016).

3 There have been three other ISSP Work Orientations modules in 1989, 2005 and 2015. Each of these modules, however, has collected slightly different data than the 1997 module. The country regions covered by the 1997 module are Germany (East and West separately), Great Britain, U.S., Hungary, Norway, Sweden, Czech Republic, Slovenia, Poland, Bulgaria, Russia, New Zealand, Canada, Japan, Israel (Jews and Arabs separately), France, Cyprus, Portugal, Denmark and Switzerland.,

on Work Orientations recorded their occupation using three-digit ISCO88 codes.

My proposed indicator of occupational automatability involves aggregating individual assessments of automation-driven job loss at the level of two-digit ISCO codes, which is the most common level of analysis in studies of the economic implications of occupational task content (e.g., Goos et al. 2014).

Hence, I combine the available three-digit and four-digit occupational data and convert them into two- digit ISCO codes. To make sure that calculated averages of the individual responses are reliable, the empirical analysis limits the sample to consider only two-digit occupations for which the aggregate automatability score is based on data from minimum 20 individual respondents. However, as a robustness check, I also repeat my baseline analyses using different thresholds for the minimum number of underlying individual responses. Similarly, I also estimate models in which occupational automatability is measured at the three- or four-digit ISCO level instead of at the two-digit level. Advantage of measuring automatability at the three- or four-digit ISCO level is that the resulting indicator is more fine-grained.

Disadvantage is not only that the two-digit level is the more common level of analysis but also that using a three- or four-digit classification leaves fewer individuals per occupational code, on average, resulting in an indicator that is more noisy. Table A.1 in Appendix A presents automatability scores for selected two-digit occupations.

II.B. Validity of Measured Occupational Automatability

The indicator of occupational automatability constructed above has much intuitive appeal. Instead of imposing a personal view on what makes a job more or less automatable, the approach harnesses the wisdom of crowds to create an indicator that is void of individual biases or other such subjectivity.

Moving beyond intuitive appeal, however, there are also different pieces of evidence testifying to the validity of the automatability indicator thus constructed.

First, gauging Table A.1, measured automatability differences have strong face validity in that occupations intuitively expected to have low/high automation risk indeed have low/high automation risk (e.g., Legislators and senior officials or Physical, mathematical, engineering science professionals vs.

Machine operators and assemblers or Drivers and mobile plant operators). For a formal test of the (construct) validity of the occupational automatability indicator, I further consider how this indicator correlates with other indicators measuring related constructs, particularly occupational routineness and complexity. Though distinct constructs, measured automation risk should correlate reasonably strongly both with differences in occupational task routineness and with differences in occupational task complexity (cf. Goos et al. 2014). Table I presents the results.

<<Insert Table I about here>>

As expected, automatability correlates strongly positively with job routineness and strongly negatively with job complexity. At the same time, correlations found are not so strong to suggest that routineness or complexity are essentially the same as automatability. Hence, the automatability indicator appears to add to common routineness and complexity indicators, capturing features of occupational task content not fully captured by either of these two job characteristics.

III. EMPIRICAL APPROACH

III.A. Empirical Model and Identification Strategy

The basic empirical model used to assess the effect of automatability of individuals’ jobs on their preferences for redistribution reads:

P β β A β β ε . (1)

In this model, Pio is the preference for redistribution of individual i working in occupation o, Ao is the automatability of occupation o, Xi is a vector of individual characteristics, and Zo is a vector of occupational characteristics other than automatability, and εio is a random error term. Per my hypothesis, I expect that β1 is statistically significantly positive, as this would indicate that individuals that are more at

risk of losing their job to automation have a stronger preference for redistribution. Because the analysis involves data that are structured hierarchically with individuals nested in occupations, I use robust standard errors that are clustered at the level of occupations.

As indicated, key challenge for valid identification of the causal effect of occupational automation risk on preferences for distribution is properly controlling for omitted variables. Two factors stand out as possible sources of an omitted variable bias. The first is a personal trait or generic preference factor that might affect both individuals’ preference for redistribution and their preference for jobs with a particular task content, which, in turn, might correlate with the degree to which these jobs are automatable. The second factor concerns an individual’s skill level, which likely affects both his/her net pecuniary gain from redistribution and his/her ability to find employment in jobs with particular task content. Notably, it seems likely that individuals with comparatively low skill levels have more difficulty finding employment in occupations with low automation risk and vice versa.

Following these two main sources of potential omitted variable bias, the main analyses include an extensive set of control variables, not least of which are various measures of individuals’ preferences.

Concerning skill level, basic control variables are years of education and educational degree, but also measures of individuals’ employment status, which, in turn, is partly a realized outcome of occupational automatability. In addition, I check results with key features of individuals’ occupation, specifically occupational task routineness and occupational task complexity, controlled for. Details on the specific variables and measures used, which vary between data sets, are presented below and in Tables A.2-A.4 in Appendix A.

The paper’s main identification strategy is to consider the effect of an individual’s indirect exposure to automation risk through the occupation of his/her spouse or partner, while controlling for the individual’s own, direct exposure to automation risk. The underlying logic is as follows. First, spousal automation risk is likely to affect an individual’s wealth prospects and hence his/her expected net pecuniary gains from government redistribution. Compared to the direct effect of the automatability of one’s own occupation, the effect of spousal automation risk is likely much weaker, but the effect may still be significant. Second,

the measure of individuals’ own occupational automation risk provides a powerful means of controlling for omitted variables, not least any unobserved differences in preferences or skill levels not yet captured by the various other control variables.

III.B. Data Sources

The main data for this paper is individual-level survey data from the ISSP modules on Social Inequality or ISSP-SI (ISSP Research Group 2014). The modules have been conducted in 1987, 1992, 1999 and 2009 and include data on respondents from 28 country regions.⁴ In addition to data on self-reported preferences for redistribution and various other important variables, these modules have recorded respondents’ occupation using the 1988 version of the ISCO. The measure of occupational automation risk applies the same occupational classification so that the individual-level ISSP-SI data can be matched to the automatability indicator constructed and validated in the previous section. The ISSP Social Inequality module(s) are commonly used to study preferences for redistribution (Corneo and Grüner 2000; Eugster et al. 2011; Kerr 2014). Dropping respondents with missing answers leaves about 45,000 individuals, depending on the model specification used. Table A.2 in Appendix A presents a description of the variables in the four ISSP Social Inequality modules used for the empirical analysis and some descriptive statistics.

Whereas the analyses using ISSP-SI data are my main analyses, as part of my robustness checks I also consider samples of individuals from two alternative data sources. The first of these alternative sources is the European Social Survey or ESS, Waves 1-7 (European Social Survey 2016). These data have been collected bi-annually during the period 2002-2014 and cover 32 mostly European and some Eurasian

4 These country regions are Australia, Austria, Bulgaria, Canada, Chile, Cyprus, Czech Republic and Slovak Republic (both also as part of former Czechoslovakia), France, (East and West) Germany, Great Britain, Hungary, Israel, Italy, Japan, Latvia, New Zealand, Norway, Philippines, Poland, Portugal, Russia, Slovak Republic, Slovenia, Spain, Sweden, Switzerland and U.S.

countries. Chief motivation for using data from the ESS is that the ESS includes interesting individual- level (control) variables not available in the ISSP Social Inequality modules. In addition, the ESS covers a different group of countries than the ISSP-SI does, which increases the international generalizability of my results.⁵ Dropping respondents with missing answers leaves almost 150,000 individuals, depending on the model specification used. Table A.3 in Appendix A presents a description of the variables in the ESS used for the empirical analysis and some descriptive statistics. The second alternative data source is the third and fourth Role of Government module of the ISSP or ISSP-ROG (ISSP Research Group 2008).

These data have been collected in 1996 and 2006 and cover 20 country regions.⁶ Main motivation for using ISSP-ROG data is that these data provide an alternative dependent variable concerning individuals’

preference for government support of declining industries. The advantage is that this preference is more specific about the preferred policy action. In addition, considering this particular preference as the dependent variable enables including measures of more generic preferences for redistribution as control variables, providing powerful means to address potential omitted variable bias. Dropping respondents with missing answers leaves about 29,000 individuals, depending on the model specification used. Table A.4 in Appendix A presents a description of the variables in the third and fourth ISSP Role of Government modules used for the empirical analysis and some descriptive statistics.

5 These country regions are Austria, Belgium, Bulgaria, Switzerland, Cyprus, Czech Republic, Germany, Denmark, Estonia, Spain, Finland, France, United Kingdom, Greece, Croatia, Hungary, Ireland, Israel, Iceland, Italy, Lithuania, Luxembourg, Netherlands, Norway, Poland, Portugal, Russia, Sweden, Slovenia, Slovakia, Turkey, and Ukraine.

6 Data from the other ISSP Role of Government modules cannot be used because these modules have not collected data on occupation. The countries covered by the third and fourth Role of Government module are Australia, Canada, Czech Republic, France, Germany, Hungary, Ireland, Israel, Japan, Latvia, New Zealand, Norway, Poland, Russia, Slovenia, Spain, Sweden, Switzerland, Great Britain and U.S.

III.C. Variables and Measures III.C.1. Dependent Variable

The main dependent variable in this paper is an individual’s preference for redistribution. The specific item from the ISSP-SI that I use to measure this preference asks respondents whether the government should reduce income differences. The first part of this item reads: “It is the responsibility of the government to reduce the differences in income between people with high incomes and those with low incomes.” Possible answers are given by the following Likert -type scale: “1, Strongly agree,” “2, Agree,”

“3, Neither agree nor disagree,” “4, Disagree,” or “5, Strongly disagree.” To facilitate interpretation, I reverse code scores on this item so that higher scores indicate a stronger preference for redistribution.

Similarly, the main analyses reported in this paper treat this item as a continuous measure of redistribution preferences. However, I obtain similar results when I estimate the empirical models using ordered probit or ordered logit techniques (detailed results available on request). As mentioned, the measure of preferences for redistribution available from the ISSP-SI data is widely used (e.g., Corneo and Grüner 2000). However, to be complete, Table A.5 in Appendix A presents some stylized evidence on the validity of this measure. If the ISSP-SI measure of preferences for redistribution is valid, we would, expect, for instance, a positive relation between this measure and measures of individuals’ attitudes towards basic income and unemployment benefits, which is confirmed by the evidence.

The ESS item measuring preferences for redistribution is highly similar to the ISSP-SI item. It asks respondents about the extent to which they agree or disagree with the statement that “the government should take measures to reduce differences in income levels.” The answer scale is again a five-point Likert-type scale: “1, Agree strongly,” “2, Agree,” “3, Neither agree nor disagree,” “4, Disagree,” or “5, Disagree strongly.” As before, I reverse code scores on this item.

Finally, as indicated, data available from the ISSP-ROG offers the opportunity to assess the effect of occupational automatability on a concrete economic policy with redistributive consequences in the form of government support of declining industries. The relevant item starts with a generic text asking about economic policies: “Here are some things the government might do for the economy. Please show which

actions you are in favor of and which you are against.” The specific policy is thereby described as

“support for declining industries to protect jobs” and answers are recorded on a five-point Likert-type scale that is reverse coded for the empirical analysis: “1, Strongly in favor of,” “2, In favor of,” “3, Neither in favor of nor against,” “4, Against,” or “5, Strongly against.”

III.C.2. Independent Variables

For most of the empirical analyses in this paper, the key independent variable is the indicator of occupational automatability constructed in the previous section. As indicated, I match this indicator to the individual-level ISSP-SI and ISSP-ROG data using two-digit codes from the 1988 version of the ISCO. I apply the same matching procedure for the analyses involving individual-level data from the ESS.

However, Waves 6 and 7 of the ESS (2012 and 2014) have not recorded occupation using ISCO88 codes but using ISCO08 codes. Hence, for these two waves I first convert ISCO08 codes into ISCO88 codes using the crosswalk provided by Ganzeboom and Treiman (2015). All the analyses include year/wave fixed effects that capture measurement error that is specific to particular waves.

As discussed, as part of my identification strategy instead of considering the effect of individuals’ own automation risk, I consider the effect of individuals’ exposure to automation risk through the occupation of their spouse or partner. All three data sources—ISSP-SI, ESS and ISSP-ROG—have collected data on the occupation of individuals’ spouses using ISCO codes. Although considering spousal occupation risk has the advantage that the automatability of individuals’ own occupation can be used as a generic control variable, the disadvantage is that the sample is reduced to individuals with a partner only. As before, matching of the occupational automatability indicator is done on the basis of two-digit codes from the 1988 version of the ISCO. Similarly, spousal occupation data from Waves 6 and 7 of the ESS are again converted from ISCO08 codes into ISCO88 codes before matching.

III.C.3. Basic Control Variables

The main empirical models that I estimate include various control variables. Some basic control variables

that I consider are sex (1=male), age and age squared. Because my data are cross-national and collected at different points in time, the basic set of controls not only includes year (or wave) dummies but also country dummies. Preferences for redistribution and occupational automatability likely correlate with education. Hence, I also control for measures of educational differences. To be exhaustive, I thereby use two different measures, one concerning years of education and one concerning level of education (e.g., no formal qualification, lowest formal qualification, et cetera). Similarly, although employment status is partly a realized outcome of occupational automation risk, all models include controls for employment status as a proxy for individual skill differences. Because prior studies find a role for religion in shaping preferences for redistribution (e.g., Basten and Betz 2013), I further include sets of dummies both for individuals’ religious denomination and for their attendance of religious services. Tables A.2-A.4 in Appendix A provide details on the measures used.

As much as possible, the control variables that I include to address potential omitted variable bias are measured at the individual level. The skill intensity of one’s occupation may, for instance, affect preferences for redistribution but this effect is controlled for by including measures of individuals’ years of education and education level. Notwithstanding, to make sure that any found effect of occupational automatability on preferences for redistribution is genuinely due to automatability I also consider two other features of individuals’ occupation, namely its routineness and its complexity. The specific measures that I use are the same measures of job routineness and job complexity considered earlier (see Table I for details). An important downside of including these two control variables is that they are available for fewer occupations than the measure of occupational automatability is. Hence, the models that control for occupational task routineness and complexity are not my preferred models.

III.B.4. Sample-Specific Control Variables

Whereas the basic control variables are independent from the sample considered, ISSP-SI, ESS or ISSP- ROG, further control variables that I use are typically only available for a specific sample.

ISSP-SI Control Variables

Relevant control variables that are specific to the ISSP-SI sample concern individuals’ economic and social status and their experience of socioeconomic mobility. Both these factors are partly a realized outcome of occupational automation risk but also important determinants of preferences of redistribution (see, for example, Alesina et al. 2018 or Corneo and Grüner 2002). I measure economic status using an item asking respondents about their social class (e.g., working class or middle class) and social status using an item asking respondents to rank their social position in society, whether they tend towards the top or towards the bottom. Concerning socioeconomic mobility, I consider individuals’ actual experience of mobility, which is measured by a set of dummies that measures how the prestige of an individual’s job compares to the prestige of the job of his/her father. In addition, I consider a measure of individuals’ self- reported beliefs about what it is needed for getting ahead in society, specifically the importance of family income. Table A.2 provides details on these items and the measures used in the analyses of the ISSP-SI sample.

ESS Control Variables

Relevant control variables that are specific to the ESS sample again concern economic status and experienced socioeconomic mobility but also two other measures that speak to the net pecuniary gains that an individual can expect from redistribution and his/her skill level. In addition, the analyses of the ESS sample consider some unique measures of individuals’ preferences and motivations. To proxy economic status I use the ESS item asking respondents about their income (measured on a 10- or 12-point scale) to create an indicator of rank income, meaning the percentile score of an individual’s income relative to other individuals in the same country. Similarly, I consider experienced socioeconomic mobility by including two sets of dummies capturing features of the socioeconomic status of individuals’

parents. The first set concerns the level of education of an individual’s father and mother respectively, while the second set concerns the employment status of the individual’s father and his/her mother when the individual was 14 years old. A further ESS-specific control variable that I consider because it likely

affect individuals’ expected benefits from government redistribution is their health status where individuals with poorer health are expected to benefit more from redistribution and therefore to have stronger preferences for redistribution. Similarly, I consider an individual’s prior unemployment experience (1=yes; 0=no), as this variable speaks to the individual’s skill level but also to the likelihood of future unemployment and hence his/her income and wealth prospects.

Concerning individuals’ preferences I include a variety of measures. The first measure is the classic left-right political self-placement scale in which respondents identify their political preferences on a spectrum that ranges from left to right (e.g., Di Tella and MacCulloch 2005). Left-right political orientation has been identified as part of a cluster of preferences involving the role of government in the economy, including preferences for redistribution (Jæger 2008; Knutsen 1995; Scheepers and Te Grotenhuis 2005). Hence, including this measure provides powerful means to rule out that there is a generic preference factor that causes individuals with a strong preference for redistribution to self-select into occupations that are highly automatable and vice versa.

The second measure is a composite index of individuals’ trust in politics, specifically trust in their country’s parliament and trust in politicians. The idea is that occupational automation risk affects individuals’ political attitudes, which then may go on to affect their preferences for redistribution. By controlling for trust in politics, however, the analysis can focus on the direct effect of automation risk on preferences for redistribution. Table A.3 provides details on the construction of the trust in politics index and the various other items and measures used in the analyses of the ESS sample.

Finally, I include a set of indicators measuring individuals’ basic values, as identified by Schwartz’s framework of universal human values, the leading framework of personal values in psychology (e.g., Schwartz 1992). The 10 basic values in this framework are universal in that they are recognized in all cultures and distinct in that they refer to different motivations. The values further form a circumplex that reflects the compatibility of each motivation with the other motivations. Specifically, values that are close in the circumplex have compatible motivations, referring to goals that can be achieved simultaneously without one necessarily coming at the expense of the other (see Figure B.1 in Appendix B). Values that

are opposite each other in the circumplex, in contrast, are not compatible and cannot be achieved simultaneously. Finally, the framework conceptualizes values as having a relative priority only and not an absolute priority. Hence, it is not possible for individuals to attach great value to everything. A short description of the 10 basic values is as follows (Schwartz et al. 2001, p. 270): Power refers to a desire for social status and prestige and for control or dominance over people and resources; Achievement refers to a desire for personal success; Hedonism refers to a desire for pleasure and enjoying life; Stimulation refers to a desire for novelty and an exciting and challenging life; Self-direction refers to a desire for being creative and independent and having freedom; Universalism refers to a desire for understanding, appreciating and tolerating others, and protection of the welfare of all people and nature; Benevolence refers to a desire for preserving and enhancing the welfare of people with whom one is in frequent personal contact; Tradition refers to a desire for respect for, commitment to, and acceptance of the customs and ideas that traditional culture or religion provide individuals; Conformity refers to a desire to restrain oneself to avoid actions likely to upset or harm others and violate social expectations or norms;

finally, Security refers to a desire for safety, harmony, and social order. Appendix B provides detailed information on the empirical operationalization of the framework of universal human values using questionnaire items included in the ESS.

ISSP-ROG Control Variables

Relevant control variables that are specific to the ISSP-ROG sample again concern different measures of individuals’ preferences. The Role of Government modules of the ISSP contain data on individuals’

preferences towards a variety of social and economic policies. Taking individuals’ preference for government support of declining industries as the dependent variable thus creates the opportunity to use measures of other policy preferences as control variables. I consider four such preferences. First and foremost, when analyzing individuals’ preference for government support of declining industries, I control for individuals’ generic preference for redistribution. The second policy preference concerns individuals’ preference for government financing of projects for new jobs. The third policy preference

concerns individuals’ preference for government spending on unemployment benefits. Finally, the analysis of the ISSP-ROG sample considers the preference for government responsibility in providing a decent living standard for the unemployed. Together, these four measures likely capture important individual differences concerning both preferences and skills that could otherwise bias the analyses using individuals’ preference for government support of declining industries as the dependent variable. Table A.4 present details on the various ISSP-ROG measures used. Meanwhile, due to missing data, analyses of the ISSP-ROG sample are unable to control for some of the factors considered in the analyses of the ISSP-SI and ESS samples, for instance socioeconomic mobility.

IV. RESULTS

IV.A. Evidence from the ISSP-SI Sample

Table II presents the baseline results. Consistent with my hypothesis, results indicate a strong positive relationship between the strength of individuals’ preferences for redistribution and their occupational automatability or automation risk (Model 1). This relationship becomes only slightly less strong when controlling for measures concerning socioeconomic mobility (Model 2) where socioeconomic mobility has the expected effect on preferences for redistribution. For example, compared to individuals with a job that has much more prestige than their father’s job, the preference for redistribution of individuals with a job that has much less prestige than their father’s job is about 0.08 standard deviations higher. Adding variables concerning economic and social status does lower the estimated coefficient for occupational automatability. However, the relationship remains highly statistically significant (Models 3 and 4).

Results are also largely the same when considering job routineness and job complexity as additional control variables (Model 5). In fact, the estimated coefficient for automation risk increases somewhat in Model 5 compared to Model 4. Possible explanation for this finding involves the criterion requiring at least 20 individual observations per occupation when constructing occupational routineness and complexity indicators, which filters out occupations that are relatively rare (see Section II). Hence,

compared to Model 4, Model 5 includes fewer occupations that are relatively rare and for which occupational automatability is measured with more measurement error. This reduction in measurement error, in turn, could shift the size of the estimated coefficient for occupational automation risk upwards. In general, the model with occupational controls is not my preferred model, however, given that this model concerns fewer observations, particularly 27 instead of 33 occupations.

<<Insert Table II about here>>

Comparing effect sizes, say, social class seems more important for understanding differences in preferences for redistribution than occupational automation risk, as evidenced by the size of the standardized coefficient (-0.097 vs. 0.054; Model 3). However, part of the effect of occupational automation risk on preferences for redistribution involves occupational automatability negatively impacting individuals’ socioeconomic status. Moreover, automatability is measured at the occupation level and not at the individual level where there is more variation.

Finally, results are robust to the use of indicators of occupational automatability constructed using fewer or more than 20 underlying individual observations (see Section II) or considering automatability measured at the three- or four-digit ISCO level instead of at the two-digit level (Table A.6 in Appendix A). The most noticeable difference is that in the models with occupational automatability measured at the three- or four-digit level (Models A3 and A4), the estimated coefficient for automation risk is lower than before. As before, possible explanation for this finding is that a more fine-grained occupational classification implies that the automatability indicator is calculated using fewer underlying individual observations per occupation on average, likely resulting in more measurement error and a corresponding downward shift in the estimated coefficient.

<<Insert Table III about here>>

Digging deeper, results indicate that the apparent effect of exposure to automation risk also occurs when considering the occupational automation risk of individuals’ spouses instead of—or, more correctly, in addition to—individuals’ own occupational automation risk (Table III). Controlling for employment status, social class or job routineness, among others, as in Table II goes a long way in ruling out that the relationship between occupational automatability and preferences for redistribution is spurious, driven by, say, unobserved differences in individuals’ skills. However, given that the relationship between exposure to automation risk and preferences for redistribution also exists in case of indirect exposure, it seems even more implausible that some unobserved differences between individuals are driving this apparent relationship. Overall, the evidence thus strongly supports a significant effect of occupational automation risk on individuals’ preferences for redistribution.

IV.B. Evidence from Other Samples and Further Robustness Checks

To extend the main analyses involving the ISSP-SI sample and provide further evidence of the robustness of the effect of occupational automatability on preferences for redistribution, Tables A.7-A.8 in Appendix A present results for analyses using the ESS sample. In all cases, results confirm the conclusion that (direct or indirect) exposure to automation risk significantly strengthens individuals’ preferences for redistribution. In addition, control variables that are unique to this sample, for example political left-right self-placement or health status, have the expected relationship with individuals’ preference for redistribution.

Turning to the ISSP-ROG sample and using a dependent variable that concerns a concrete policy action with redistributive consequences, results again support the earlier conclusion that occupational automation risk affects individuals’ preferences for redistributive government policy (Table A.9 in Appendix A). Also with, among others, individual differences in the generic preference for redistribution and for government financing of projects for new jobs controlled for, there is a significant relationship between (direct or indirect) exposure to automation risk and the strength of individuals’ preference for government support of declining industries.

V. CONCLUSION

This paper shows that individuals in occupations that are more at risk of losing their job to automation have stronger preferences for redistribution. This result is supported by evidence from three different large-scale cross-country survey data sets and extends to preferences for a concrete policy action with redistributive consequences in the form of government support for declining industries. Furthermore, findings are robust to the inclusion of a large variety of control variables, including some unique measures of individuals’ preferences and variables such as socioeconomic status that are partly a realized outcome of occupational automation risk. Finally, the effect of automation risk on preferences for redistribution is also found when considering indirect exposure to automation risk through the occupation of one’s spouse or partner. Hence, results do not appear driven by unobserved individual differences in preferences or skill level.

Recent years have seen an increasingly intense debate about the possible disruptive effects of technological advances in robotics and artificial intelligence algorithms, among others. A most prominent issue thereby involves the distributional effects of technological change and the concern that some groups in society will not only lose from automation in relative terms but possibly also in absolute terms.

However, as this paper shows, automation can have important societal implications even when the objective threat of automation causing unemployment and income losses is limited. The reason is that fear and anxiety about automation alone are sufficient to radically alter people’s political attitudes and increase the pressure on governments to take action.

Appendix A

<<Insert Tables A.1-A.9 here>>

Appendix B

This appendix describes the operationalization of the framework of universal human values using the Portrait Values Questionnaire or PVQ developed by Schwartz and collaborators (Schwartz et al. 2001).

The PVQ consists of 21 “portraits” that ask individuals to describe themselves. The generic item text reads: “Now I will briefly describe some people. Please listen to each description and tell me how much each person is or is not like you.” Answers can range from “1, Very much like me” to “6, Not like me at all.” Table B.1 describes the 21 portraits and how they relate to the 10 basic values and motivations recognized in the universal human values framework.

<<Insert Table B.1 about here>>

As mentioned in the main text, in the universal human values framework, values only have a relative priority and not an absolute priority. Practically, getting from absolute scores on the PVQ items to relative priorities, requires that we ipsatize all individual ratings (Schwartz et al. 2001). To do so, we first need to calculate individuals’ average score on the 21 items and then subtract this average score from individuals’

scores on each item. This renders item scores that are standardized to indicate how weakly or strongly an individual scores on a particular value item compared to his/her score on other value items. Moreover, given that answers originally come on a 1 to 6 scale, ipsatized item scores can, in principle, vary from -5 to +5. Finally, ipsatized scores on the 21 items can be used to construct individual ratings on the 10 basic human values, following the mapping presented in Table B.1. In this case, ipsatized item scores are simply added and divided by the number of items. Theoretically, the range of possible scores on the resulting values measures can again range from -5 to +5. Figure B.1 present a graphical description of the interrelationship between the 10 distinct basic values in the framework of universal human values.

<<Insert Figure B.1 about here>>

REFERENCES

Acemoglu, Daron, and David Autor. 2011. Skills, Tasks and Technologies: Implications for Employment and Earnings. In: Orley Ashenfelter, and David Card, Eds. Handbook of Labor Economics.

Amsterdam: Elsevier, 1043-171.

____, and Pascual Restrepo. 2017. Robots and jobs: Evidence from US labor markets. NBER Working Paper No. 23285. Cambridge, MA: National Bureau of Economic Research.

Akst, Daniel. 2014. What Can We Learn from Past Anxiety over Automation? Wilson Quarterly (Summer).

Alesina, Albert, and Paola Giuliano. 2011. Preferences for redistribution. In: Jess Benhabib, Alberto Bisin, and Matthew Jackson, Eds. Handbook of Social Economics. Elsevier: Amsterdam, 93-131.

____, Stefanie Stantcheva, and Edoardo Teso. 2018. Intergenerational mobility and preferences for redistribution. American Economic Review 108: 521-54.

Ashok, Vivekinan, Ilyana Kuziemko, and Ebonya Washington. 2015. Support for Redistribution in an Age of Rising Inequality: New Stylized Facts and Some Tentative Explanations. Brookings Papers on Economic Activity: 367-433

Autor, David H. 2015. Why Are There Still So Many Jobs? The History and Future of Workplace Automation. Journal of Economic Perspectives 29(3): 3-30.

____, and David Dorn. 2013. The Growth of Low-Skill Service Jobs and the Polarization of the US Labor Market. American Economic Review 103: 1553-97.

____, David Dorn, and Gordon H. Hanson. 2015. Untangling Trade and Technology: Evidence from Local Labor Markets. Economic Journal 125: 621-46.

____, Frank Levy, and Richard J. Murnane. 2003. The Skill Content of Recent Technological Change: An Empirical Exploration. Quarterly Journal of Economics 118: 1279-333.

Basten, Christoph, and Frank Betz. 2013. Beyond work ethic: Religion, individual, and political

preferences. American Economic Journal: Economic Policy 5: 67-91.

Bénabou, Roland, and Efe A. Ok. 2001. Social Mobility and the Demand for Redistribution: The POUM Hypothesis. Quarterly Journal of Economics 116: 447-87.

Blinder, Alan S., and Alan B. Krueger. 2013. Alternative measures of offshorability: a survey approach.

Journal of Labor Economics 31: S97-128.

Brynjolfsson, Erik, and Andrew McAfee. 2014. The Second Machine Age: Work, Progress, and Prosperity in a Time of Brilliant Technologies. New York and London: W.W. Norton & Company.

Ciccone, Antonio, and Elias Papaioannou. 2016. Estimating cross-industry cross-country interaction models using benchmark industry characteristics. NBER Working Paper No. 22368. Cambridge, MA:

National Bureau of Economic Research.

Clark, Lee Anna, and David Watson. 1995. Constructing validity: Basic issues in objective scale development. Psychological Assessment 7: 309-19.

Corneo, Giacomo, and Hans Peter Grüner. 2002. Social limits to redistribution. American Economic Review 90: 1491-507.

____, and ____. 2002. Individual Preferences for Political Redistribution. Journal of Public Economics 83: 83-107.

Costinot, Arnaud, Lindsay Oldenski, and James Rauch. 2011. Adaptation and the boundary of multinational firms. Review of Economics and Statistics 93: 298-308.

Dahlberg, Matz, Karin Edmark, and Heléne Lundqvist. 2012. Ethnic Diversity and Preferences for Redistribution. Journal of Political Economy 120: 41-76.

Di Tella, Rafael, and Robert MacCulloch. 2005. Partisan social happiness. Review of Economic Studies 72: 367-93.

Etzioni, Amitai. 2017. Job Collapse on the Road to New Athens. Challenge 60: 327-46.

Eugster, Beatrix, Rafael Lalive, Andreas Steinhauer, and Josef Zweimüller. 2011. The demand for social insurance: does culture matter? Economic Journal 121: F413-48.

European Social Survey. 2016. European Social Survey Cumulative File, ESS 1-7. Data file edition 1.0.

NSD - Norwegian Centre for Research Data, Norway - Data Archive and distributor of ESS data for ESS ERIC.

European Foundation for the Improvement of Living and Working Conditions. 2018. European Working Conditions Survey Integrated Data File, 1991-2015. [data collection]. 7th Edition. UK Data Service.

SN: 7363, http://doi.org/10.5255/UKDA-SN-7363-7.

Fong, Christina. 2001. Social preferences, self-interest, and the demand for redistribution. Journal of Public Economics 82: 225-46.

Ford, Martin. 2015. Rise of the Robots: Technology and the Threat of a Jobless Future. New York: Basic Books.

Frey, Carl Benedikt, and Michael A. Osborne. 2017. The future of employment: How susceptible are jobs to computerisation? Technological Forecasting and Social Change 114: 254-80.

Ganzeboom, Harry B.G., and Donald J Treiman. 2015. International Stratification and Mobility File:

Conversion Tools. Amsterdam: Department of Social Research Methodology, http://www.harryganzeboom.nl/ismf/index.htm.

Goos, Maarten, Alan Manning, and Anna Salomons. 2014. Explaining Job Polarization: Routine-Biased Technological Change and Offshoring. American Economic Review 104: 2509-26.

Hobsbawm, Eric J. 1952. The machine breakers. Past & Present 1: 57-70.

ISSP Research Group. 1999. International Social Survey Programme: Work Orientations II - ISSP 1997.

GESIS Data Archive, Cologne. ZA3090 Data file Version 1.0.0, doi: 10.4232/1.3090.

____. 2008. International Social Survey Programme: Role of Government I-IV - ISSP 1985-1990-1996- 2006. GESIS Data Archive, Cologne. ZA4747 Data file Version 1.0.0, doi: 10.4232/1.4747.

____. 2014. International Social Survey Programme: Social Inequality I-IV - ISSP 1987-1992-1999-2009.

GESIS Data Archive, Cologne. ZA5890 Data file Version 1.0.0, doi: 10.4232/1.11911.

Jæger, Mads Meier. 2008. Does left–right orientation have a causal effect on support for redistribution?

Causal analysis with cross-sectional data using instrumental variables. International Journal of Public Opinion Research 20: 363-74.

Kerr, William R. 2014. Income inequality and social preferences for redistribution and compensation differentials. Journal of Monetary Economics 66: 62-78.

Keynes, John Maynard. 1930. Economic Possibilities for our Grandchildren. In: John Maynard Keynes.

Essays in Persuasion. New York: Classic House Books, 358-73.

Krugman, Paul. 2012. Rise of the Robots. New York Times. December 8.

Kuziemko, Ilyana, Michael I. Norton, Emmanuel Saez, and Stefanie Stantcheva. 2015. How Elastic Are Preferences for Redistribution? Evidence from Randomized Survey Experiments. American Economic Review 105: 1478-508.

Luttmer, Erzo F.P. 2001. Group Loyalty and the Taste for Redistribution. Journal of Political Economy 109: 500-28.

____, and Monica Singhal. 2011. Culture, Context, and the Taste for Redistribution. American Economic Journal: Economic Policy 3: 157-79.

Marx, Karl. 1961 [1867]. Capital. Moscow: Foreign Languages Publishing House.

McKinsey. 2017. A future that works: Automation, employment, and productivity. McKinsey Global Institute.

Meltzer, Alan, and Scott Richards. 1981. A Rational Theory of the Size of Government. Journal of Political Economy 89: 914-27.

Mokyr, Joel, Chris Vickers, and Nicolas L. Ziebarth. 2015. The history of technological anxiety and the future of economic growth: Is this time different? Journal of Economic Perspectives 29(3): 31-50.

Piketty, Thomas. 1995. Social Mobility and Redistributive Politics. Quarterly Journal of Economics 110:

551-84.

Ravallion, Martin, and Michael Lokshin. 2000. Who wants to redistribute? The tunnel effect in 1990 Russia. Journal of Public Economics 76: 87-104.

Roberts, Kevin W.S. 1977. Voting over Income Tax Schedules. Journal of Public Economics 8: 329-40.

Romer, Thomas. 1975. Individual Welfare, Majority Voting and the Properties of the Linear Income Tax.

Journal of Public Economics 4: 163-8.

Sachs, Jeffrey D., and Laurence J. Kotlikoff. 2012. Smart machines and long-term misery. NBER Working Paper No. 18629. Cambridge, MA: National Bureau of Economic Research.

Sale, Kirkpatrick. 1995. Rebels against the future: the Luddites and their war on the Industrial Revolution:

lessons for the computer age. Reading, MA: Addison-Wesley.

Scheepers, Peer, and Manfred Te Grotenhuis. 2005. Who cares for the poor in Europe? Micro and macro determinants for alleviating poverty in 15 European countries. European Sociological Review 21: 453- 65.

Schwartz, Shalom H. 1992. Universals in the content and structure of values: Theoretical advances and empirical tests in 20 countries. Advances in Experimental Social Psychology 25: 1-65.

____, Gila Melech, Arielle Lehmann, Steven Burgess, Mari Harris, and Vicki Owens 2001. Extending the cross-cultural validity of the theory of basic human values with a different method of measurement.

Journal of Cross-Cultural Psychology 32: 519-42.

Srnicek, Nick, and Alex Williams. 2015. Inventing the future: Postcapitalism and a world without work.

New York: Verso.

Summers, Lawrence H. 2013. Economic Possibilities for Our Children. The 2013 Martin Feldstein Lecture. NBER Reporter 4, 1-6.

Table I. Construct Validity of Occupational Automatability.

Correlations for 27 occupations Automatability Routineness

Automatability 1

Routineness 0.542 1

Complexity -0.623 -0.773

Notes: Indicators of occupational task routineness and complexity are constructed to concern occupations at the two-digit level of the 1988 version of the ISCO. Data on job routineness and 1988 ISCO codes are available from Waves 2 and 5 of the European Social Survey or ESS (European Social Survey 2016). I measure routineness using the ESS item asking respondents to indicate how much the statement “there is a lot variety in my work” applies to their job. Respondents can respond using the following answer scale:

“1, Not at all true,” “2, A little true,” “3, Quite true,” or “4, Very true.” The occupational routineness indicator is constructed by reverse coding individual answers on this item and aggregating them at the two-digit ISCO level. Data on job routineness and complexity and 1988 ISCO codes are available from Waves 3-6 of the European Working Conditions Survey or EWCS (European Foundation for the Improvement of Living and Working Conditions 2018). I measure job complexity using the EWCS item asking respondents whether their generally job involves complex tasks (1) or not (0). The occupational complexity indicator is constructed by aggregating individual answers on this item at the two-digit ISCO level. Dropping occupations with fewer than 20 underlying individual responses for both the routineness and the complexity indicator leaves 27 two-digit occupations.

Table II. Occupational Automatability and Preferences for Redistribution.

Dependent = preference for

redistribution 1 2 3 4 5

Automation risk 0.077 (0.018)

[0.000] 0.074 (0.018)

[0.000] 0.054 (0.014)

[0.001] 0.054 (0.014)

[0.001] 0.057 (0.016) [0.002]

Sex (1=Male) -0.129 (0.022)

[0.000] -0.131 (0.022)

[0.000] -0.126 (0.018)

[0.000] -0.130 (0.018)

[0.000] -0.128 (0.021) [0.000]

Years of education -0.011 (0.014)

[0.408] -0.011 (0.014)

[0.424] 0.012 (0.013)

[0.353] 0.012 (0.013)

[0.370 0.015 (0.013) [0.248]

Belief about importance of family

wealth for getting ahead - 0.066 (0.006)

[0.000] - 0.057 (0.006)

[0.000] 0.057 (0.006) [0.000]

Rank of social class - - -0.097 (0.007)

[0.000] -0.095 (0.007)

[0.000] -0.092 (0.007) [0.000]

Rank of top-bottom self-placement - - -0.094 (0.006)

[0.000] -0.091 (0.006)

[0.000] -0.091 (0.006) [0.000]

Job routineness - - - - 0.002 (0.028)

[0.938]

Job complexity - - - - -0.009 (0.027)

[0.745]

Dummies for occupational prestige

relative to father No Yes No Yes Yes

Dummies for employment status Yes Yes Yes Yes Yes

Dummies for education level Yes Yes Yes Yes Yes

Dummies for religious denomination Yes Yes Yes Yes Yes

Dummies for religious attendance Yes Yes Yes Yes Yes

Age and age² Yes Yes Yes Yes Yes

Dummies for country Yes Yes Yes Yes Yes

Dummies for year Yes Yes Yes Yes Yes

No. of occupations 33 33 33 33 27

No. of individuals 44,935 44,935 44,935 44,935 44,493

R² 0.1987 0.2033 0.2192 0.2225 0.2229

Notes: Results obtained using the ISSP-SI sample. Standard errors (in parentheses) are robust standard errors that are clustered at the level of occupations. P-values are in square brackets. The dependent variable and continuous independent variables are standardized to have a mean of 0 and a standard deviation of 1. To save space, the table presents a selection of coefficients, standard errors and p-values but complete results are available on request.

Table III. Spousal Automation Risk and Preferences for Redistribution.

Dependent = preference for

redistribution 6 7 8

Spousal automation risk 0.034 (0.012)

[0.006] 0.028 (0.012)

[0.024] 0.026 (0.010) [0.017]

Automation risk - 0.055 (0.008)

[0.000] 0.055 (0.009) [0.000]

Spousal job routineness - - 0.022 (0.014)

[0.130]

Spousal job complexity - - 0.003 (0.014)

[0.801]

Other control variables Yes Yes Yes

No. of spousal occupations 33 33 27

No. of individuals 29,303 25,685 25,376

R² 0.2374 0.2490 0.2515

Notes: Results obtained using the ISSP-SI sample. Standard errors (in parentheses) are robust standard errors that are clustered at the level of spousal occupations. P-values are in square brackets. The dependent variable and continuous independent variables are standardized to have a mean of 0 and a standard deviation of 1. Other control variables are years of education, sex, belief about importance of family wealth for getting ahead, rank of social class, rank of top-bottom self-placement, dummies for employment status, dummies for religious denomination, dummies for religious attendance, dummies for occupational prestige relative to father, dummies for education level, age, age squared, dummies for country and dummies for year (see Model 4 in Table II). To save space, the table presents a selection of coefficients, standard errors and p-values but complete results are available on request.

Table A.1. Automatability of Selected Occupations.

Occupation (two-digit ISCO88 code in square brackets) Automatability score

Legislators and senior officials [11] 3.24

Physical, mathematical and engineering science professionals [21] 3.35

Corporate managers [12] 3.53

General managers [13] 3.66

Customer services clerks [42] 3.83

Metal, machinery and related trades workers [72] 3.84

Office clerks [41] 3.84

Machine operators and assemblers [82] 3.91

Drivers and mobile-plant operators [83] 3.96

Notes: Source is own calculations based on data from the 1997 ISSP Work Orientations module (ISSP Research Group 1999).